NATURAT. ORDERS. 117 



i7icompIetnw, ow\y ouG circular memljranc exists, with the 

 unilateral nuliment of the second. 



The rudiment of the inferior membrane in this si)ecies 

 points out the relation between the ap])arently anomalous 

 appendage of the calyx in Tristemma, and the ciliated 

 scales irregularly scattered over its whole surface in 

 Osbeckia ; the analogy being established by the interme- 

 diate structure of an unpublished plant of this order fmni 

 Sierra Leone, in Sir Joseph Banks's herbarium, in which 

 the nearly similar squama3, though distinct, are disposed 

 in a single complete circle ; and by Melastoma octandra of 

 Linnaeus, in which they are only four in number, and alter- 

 nate with the proper divisions of the calyx. 



The two species here referred, though improperly, to 

 Rhexia, agree with a considerable part of the species pub- 

 lished in the monograph of that genus by M. Bonpland, 

 and with some other genera of the order, in the peculiar 

 manner in which the ovarium is connected with the tube 

 of the calyx. This cohesion, instead of extending uni- 

 formly over the whole surface, is limited to ten longitudinal 

 equidistant lines or membranous processes, apparently 

 originating from the surface of the ovarium ; tlie inter- 

 stices, w^hich are tubular, and gradually narrowdng towards 

 the base, being entirely free. 



The function of these tubular interstices is as remark- 

 able as their existence. 



Li Melastomaceae, before the expansion of the corolla, the 

 tops of the filaments are inflected, and the antherae are 

 pendulous and parallel to the lower or erect portion of the 

 filament ; their tips reaching, either to the line of complete 

 cohesion between the calyx and ovarium, Avhere that exists ; 

 or, where this cohesion is partial, and such as I have now [i^>6 

 described, being lodged in the tubular interstices ; their 

 points extending to the base of the ovarium. Lrom these 

 sheaths, to which they are exactly adapted, the anthera? 

 seem to be disengaged in consequence of the unequal 

 growth of the diflerent parts of the filament ; the inllectcd 

 portion ceasing to increase in length at an early ])criod, 

 while that below the curvature continues to elongate con- 



